Research On Environmental Awareness And Path Planning Of Structured Road Based On Lidar

In recent years,the number of automobiles in my country has grown rapidly.While automobiles have brought convenience to people’s lives,they have also brought many problems,such as traffic accidents and traffic congestion.Smart vehicles have gradually been paid attention to by the public as effective solutions to these problems.As a product of the development of science and technology,intelligent vehicles can improve the situation of traffic congestion and ensure travel safety.Accurate and real-time perception of road environment is the premise for intelligent vehicle to realize autonomous driving.Only by fully grasping the surrounding environment information can intelligent vehicle plan a scientific and reasonable driving route and provide guidance.The subject of this paper is from the National Natural Science Foundation Project "Research on Depth Level Perception and Understanding of Complex Dynamic Environment of Intelligent Vehicles"(Project No.:U1864204).Combined with the Lidar sensor,the research on structured road environment perception with roadside and obstacle avoidance path planning method is carried out from the following aspects:(1)The structured road boundary is extracted based on the lidar point cloud.In order to eliminate the influence of vehicle movement and environment on point cloud data collection,the cumulative error of point cloud data coordinates was corrected by establishing a lidar movement model.Based on the distribution characteristics of the ground point cloud,two ground models were established,and the ground point segmentation was completed by improving the RANSAC algorithm.Based on the three features of the neighborhood relationship in the point cloud space,the road edge was initially extracted by establishing the road boundary point feature model,and then the adaptive direction search algorithm was used to "repair" the missing boundary segment.(2)Based on obstacle point cloud samples,obstacle clustering detection and dynamic obstacle movement information extraction are carried out.In order to overcome the shortcomings of traditional clustering algorithms that rely too much on the initial clustering center and distance threshold,the European clustering algorithm is improved according to the distribution characteristics of the obstacle point cloud,and the obstacle clustering detection is completed.Two square box models are used to model the obstacle parameters,and then the constructed obstacle multi-feature model is introduced into the adjacent frame point cloud data association matching algorithm,and the Kalman filter is used to estimate and update the target motion state.(3)The improved artificial potential field method is adopted to realize local path planning.An improved method is proposed to solve the problems of local minimum and unreachable target existing in the traditional artificial potential field method.Based on the road environment perception results,a multi-constraint model including the road boundary constraint and the steering mechanism constraint was constructed.The model was integrated into the improved artificial potential field method to construct the resultant potential field model based on the multi-constraint model,and the local path planning of the vehicle was realized.Finally,the simulation experiment is set for analysis.(4)This paper collects real vehicle data based on the built intelligent vehicle test platform,builds the point cloud processing algorithm framework based on the ROS system,and uses the collected data to verify and analyze the road environment perception method in this paper.The experimental results show that the road environment perception method proposed in this paper is effective and reliable.